Tuberculosis (TB)/HIV co-infection significantly changes the natural history of both diseases. Proper comprehension and clinical management of co-infected TB/HIV patients is still a challenge, particularly in places like Brazil, where both types of infection are prevalent. Evaluate the frequency of the clinical forms of TB in HIV-infected patients; correlate the clinical forms of TB with the level of immunodeficiency; evaluate the response to therapy with different regimens for the treatment of TB; identify potential prognostic factors in TB/HIV patients. The following data were collected at the beginning of the study: medical history, epidemiological background, physical examination, and laboratory evaluation (complete blood cell count, T lymphocyte subsets, viral load and tuberculin test). Monthly clinical follow-up was performed, with attention to adverse reactions to tuberculostatic drugs. TCD4+/CD8+ lymphocyte counts and quantification of the viral load were performed after 2, 4, 6, 10 and 15 months of follow-up. The study population consisted of 78 patients (45 males and 33 females) and their mean age was 36.4 +/- 7.9 years The mean TCD4+ count values were higher in patients with the cavitary pulmonary form and lower in patients with disseminated forms. There were no significant differences in the mean TCD8+ cells counts . in the different clinical forms of TB. However, the mean laboratory values for hemoglobin, hematocrit and leucocytes at study entry did differ significantly among the various clinical forms of TB. At the end of the trial, the Tb recovery rate was of 78%, with four cases (5%) of treatment failure, eight (11%) treatment discontinuations and five deaths (6.4%). The highest rate of treatment failure (75%) was observed among patients with the disseminated form. Lower TCD4+ mean values were observed in cases of treatment failure and death. There was a correlation between the TCD4+ cell values and the TB outcome at the six time points. TCD8+ (cells/mm(3)) mean values assessed at the six time points in relation to the TB outcome indicated (non-significantly) lower values in patients who progressed to treatment failure. Considering the different TB outcomes, there was a significant correlation between TCD8+ values at the first and third assessments. Lower mean values of hemoglobin, hematocrit, platelet and leukocytes were observed among the cases of treatment failure than in patients who recovered. The variables hemoglobin, hematocrit, leukocytes and platelets were significantly different among the groups. The pulmonary forms of TB were most frequent in HIV infected patients; the extrapulmonary, associated and disseminated forms were predominantly seen in patients with a more severe level of immunosuppression. The TB recovery rate in HIV infected patients was similar to the expected rate in non-infected individuals. The best results were obtained when regimens containing rifampicin were used. Pancytopenia and low levels of TCD4+ and TCD8+ lymphocytes at the initial time point of the study were correlated with an unfavorable outcome of TB, and therefore they can be considered potential prognostic factors. However, the of TCD8+ lymphocyte values were the most important variable assessed.